How Nvidia trained Nemotron, better agents, and more #31
Towards AI
Making AI accessible to all with our courses, blogs, tutorials, books & community.
Nvidia’s secret was… synthetic data + weak-to-strong alignment.
Good morning, AI enthusiasts!
We are excited to announce that ‘Building LLMs for Production’ is now also available to readers across the globe on the O-Reilly learning platform. But that’s not all. We are also working on more exciting collaborations with O’Reilly to bring even more value and resources to our community (we will share more about this soon!).
For over 45 years, O'Reilly has been one of the biggest platforms for providing comprehensive learning resources. It offers exclusive live training, interactive learning experiences, certification programs, books, videos, and more.
If you are a subscriber of the platform, you can read it directly on the O’Reilly learning platform or sign up for a 10-day free trial to access the book .?
If you are like us and prefer a physical book, you can also find it as a paperback, high-quality colored hardcover on Amazon . If you have already grabbed your copy, help fellow AI community members discover the book by leaving a review on Amazon . ?
What’s AI Weekly
In today’s video, I dive into key learnings from Nvidia’s Nemotron family of models and insights for training an LLM using synthetic data. Training large language models is such a massive challenge due to the enormous need for high-quality data. But getting that data is incredibly tough. While many people have tried to solve this problem in various ways, synthetic data is one of the most promising approaches. It’s less expensive than other methods but has a major drawback: the lack of diversity. Recently, Nvidia’s new LLMs from their Nemotron family of models have addressed this issue. They’ve shared a pipeline for generating synthetic data that’s used for training and refining large language models (LLMs). Watch the video or read the article version !?
— Louis-Fran?ois Bouchard , Towards AI Co-founder & Head of Community
This issue is brought to you thanks to Ai4 :?
Join the industry’s leading AI conference - free passes available!
Ai4, the world’s largest gathering of artificial intelligence leaders in business, is coming to Las Vegas - August 12-14, 2024. Join 4500+ attendees, 350+ speakers, and 150+ AI exhibitors from 75+ countries at the epicenter of AI innovation.
Don’t wait - passes are going fast. Apply today for a complimentary pass, or register now for 41% off final prices.
Register now !?
Learn AI Together Community section!
Featured Community post from the Discord
Craenius just launched a demo of their latest agent, NotDevin. Their experiments show that NotDevin was able to replace Devin and Google's Project IDX. You can sign up here to get on the waitlist and support a fellow community member. Share your feedback in the Discord thread !?
AI poll of the week!
For everyone planning to buy the book, now is a great time! We now have it as an e-book, paperback, hardcover on Amazon , and the O’Reilly learning platform. For those who don’t like books, do you all want a more bite-sized version of the most important takeaways? Tell us in the Discord thread !?
领英推荐
Collaboration Opportunities?
The Learn AI Together Discord community is flooding with collaboration opportunities. If you are excited to dive into applied AI, want a study partner, or even want to find a partner for your passion project, join the collaboration channel ! Keep an eye on this section, too—we share cool opportunities every week!?
1. Tanishk0619 is looking for a couple of people to learn ML with and be accountability partners. If you are also looking for a disciplined learning journey, contact him in the thread !
2.? Nitin01652 is pursuing deep reinforcement learning courses from HuggingFace. He is looking for partners to discuss assignments and share resources for the next courses on other topics. If you want to try it, connect with him in the thread !?
3. Baadror is starting his LLM learning journey with hands-on projects. If you want to start learning and are looking for other learners too, reach out to him in the thread !?
Meme of the week!
Meme shared by rucha8062
TAI Curated section
Article of the week
Inspired by the Kolmogorov-Arnold representation theorem, KANs emerge as promising alternatives to Multi-Layer Perceptrons (MLPs). Unlike traditional neural networks, KANs place activation functions along the connections between nodes, not at the nodes themselves. This innovative approach opens doors for further enhancing deep learning models that heavily rely on MLPs. The goal of this article is to give some basic understanding of KAN and explore the parts or building blocks of KAN in this Article.
Our must-read articles
From e-commerce to customer support, all businesses require some kind of NER model to process large amounts of texts from users. Businesses require NER models to extract relevant and important entities from text. This article explains how to build NER from scratch.
Enterprises absolutely need control of things like logging, monitoring, and security while also striving to integrate AI into their established infrastructure. Going for in-house manufacturing might not be feasible as it requires specialized knowledge, tools, and resources. This is when NVIDIA NIM comes into the picture; explore more in this article.
In this insightful article, Jan Werth dives into stable face-mask detection using an adapted eye cascader. The article explains how the adapted eye cascader works, providing step-by-step details on detecting eyes and creating face-bounding boxes.
If you are interested in publishing with Towards AI, check our guidelines and sign up . We will publish your work to our network if it meets our editorial policies and standards.
AI & ML Innovator | Transforming Data into Revenue | Expert in Building Scalable ML Solutions | Ex-Microsoft
4 个月This week in the AI community has been exciting! We've seen a great new AI agent introduced, which is getting a lot of attention. There are also some interesting opportunities for collaboration focused on LLMs, which could lead to some fantastic advancements. The articles on KAN and NER models were insightful and valuable for anyone working in the field. It's great to see so much happening and to be part of such a dynamic community. Thanks for sharing these updates!